imageflowtech/spectral_embedding_gcn

Spectral Graph Convolution for Infant Brain segmentation and age prediction

Spectral Graph Convolution for Infant Brain segmentation and age prediction

The code used for cortical surface parcellation and age regression (birth age, scan age)

Spectral embedding

The aligned spectral embedding for brain surfaces can be computed using the following repository

https://github.com/kharitz/aligned_spectral_embedding.git

Dataset

A "dataset" folder with the aligned spectral embedding is taken as input for brain surface analyis

Training

The model can be trained using below command:
To train parcellation/segmentation newtork

python3 main_train_parcellation.py --config ./parcellation/config/config_train_parcellation.json --gpu 0

To train birth age regression newtork

python3 main_train_birth_regression.py --config ./parcellation/config/config_train_birth_regression.json --gpu 0

To train scan age regression newtork

python3 main_train_scan_regression.py --config ./parcellation/config/config_train_scan_regression.json --gpu 0